Have we reached peak data in pro football?

Stuart Grant

When discussing the use of applied statistics or Moneyball in pro football, the conversation eventually turns to the intangibles of star players that appear immeasurable. Pro football heaps glorious accolades on its winners and stars usually with reference to wins and titles as opposed to numerical milestones and statistics. Moneyball has yet to bring about a paradigm shift across the NFL in the same way that the Oakland A’s success changed MLB.

The explosive success of fantasy football, however, has changed the way fans watch the game. This phenomenon has led to the casual fan becoming an “internet general manager” of their fantasy team. A good manager must keep his team equipped with productive offensive players. The requisite statistics now require watching if an owner wants to be able to walk by the watercooler at work without being catcalled for mismanaging their fantasy roster.

The free agency era created a robust market in veteran players. As the fantasy football era means that everything that happens on the field is measured, a new vocabulary has been created. When an offensive skill position player (Running Back, Wide Receiver, Tight End) changes teams or retires, their statistics like receptions, carries, targets and touches are described as being “vacated” by their departure. The implication being that these touches must be replaced if the offense is to perform at the same level.

In the NFL, quarterback is king. The league has had a passer rating system since 1973. As the game has liberalized rules to favor passing, modern day quarterbacks dwarf their counterparts of yesteryear in offensive output. Passer rating is comprised of completion percentage, yards per attempt, touchdown percentage and interception percentage.

Hall of Famer and two time Super Bowl winner, John Elway, had a career rating of 79.9 which was much lower than less decorated QBs of his era like Neil Lomax and Danny White. The passer rating system only measures passing, not overall effectiveness.

Moneyball has been with us for almost twenty years. In a tradition driven sport like baseball, it was revolutionary yet baseball has always been a more statistically driven game than football, basketball and hockey.

Moneyball’s methods have only recently been applied to pro football. As a copycat league, every NFL team now has a statistical analysis unit. What doesn’t get discussed is how much statistical analysis plays into on-field football decisions and play calling.

NFL Owners and Moneyball

If the details emerging from the firing of Philadelphia Eagles head coach Doug Pederson are any indication, team owners are banking heavily on Moneyball approaches. Depending on the philosophical bent and level of involvement of the owner, a coach’s failure to adhere to statistical probabilities in play calling could cost him his job.

This raises troubling scenarios for prospective coaches working for egomaniac owners. Raider fans recall the interventionist ways of the late owner, Al Davis, calling in plays from his luxury box. His constant meddling undermined coaches, created dysfunction and tarnished his legacy. It looks like Cowboys owner Jerry Jones is following suit.

Jeffrey Lurie’s questioning Super Bowl winning Head Coach Doug Pederson about play calling underscores the very real possibility of an egomaniac owner calling his own plays with scripts provided by his statistical analysis unit. Time will tell whether new Eagles head coach Nick Siriani is anything more than an errand boy cheerleader.

Air Yards

Defined as the distance the ball travels past the line of scrimmage before it is caught.

IAY — Intended air yards — Air yards on all pass attempts, whether completed or incomplete

IAY/PA — Intended air yards per pass attempt — Average depth of target, whether completed or not

CAY — Completed air yards — Total yards completed passes traveled in the air past the line of scrimmage before being caught

CAY/Cmp. — Completed air yards per completion — yards the ball traveled in the air past the line of scrimmage prior to a completion

CAY/PA — Completed air yards per pass attempt — Air yards (on completed passes) per pass attempt

While I understand the value of tabulating Completed Air Yards when calculating plays with a “yards after the catch” factor, the use of Intended Air Yards stats raises questions.

If all pass attempts are treated equally, a Hail Mary or throwaway to avoid a sack or kill the clock will be captured in this data. What about adverse weather games and garbage time stats? Will those have the effect of decreasing the perceived effectiveness of a QB?

Expected Points (Added)

This metric was arrived at by compiling every down and distance situation in league history and factoring in next points scored. Every down and distance situation at a specific field position is assigned a point value. Field positions on the opponent’s side of the 50 yard line are given greater value assuming that the closer a team is to the goal line, the more difficult it is to gain yards.

A team with the ball at midfield on first and ten is in a situation with an Expected Point Value of +2.0. If they gain five yards for a second down and five at the opponents 45 yard line this would result in an expected points value of +2.1. The gain of +0.1 points is expressed as Expected Points Added.

Conversely, if the Quarterback got sacked for a five yard loss, his team would effect an Expected Points value of +1.2 moving their possession back to their own 45 yard line. That 5 yard loss play would accrue -0.8 Expected Points Added.

Plays with turnovers also carry related values. Assume that our offense had the ball on second down and five at the opponents 45 yard line for a possession valued at +2.2 Expected Points. If they fumble, that play would give their opponent possession at their own 45 yard line for a possession with an Expected Point Value of +2.1. The Expected Point Value of the play for the team that lost the ball is minus 4.3.

The creators of the Expected Points metric were aware that a win by one point is worth the same to the victor as a win by 30 points. For Expected Points values to be useful they must be of relative equal value without time being a factor. In order to eliminate situational anomalies like teams killing the clock in “garbage time”, the creators only used samples from games with a ten point difference or less and only from the first and third quarters.

In other words, in creating this metric they didn’t use samples from the end of the first or second half when close games are won and lost. Other than determining whether runs or passes would have a greater chance of gaining yards, the use of Expected Points is debatable. Fans know that defenses play differently in crunch time than they do when the game is not yet on the line. If stats are not taken from situations where games are decided, how useful can they be?

One problem I have with Air Yards and Extra Points Added is that they look at plays in isolation instead of being part of a drive. Momentum is important in football. When play calling decisions are questioned in press conferences, coaches’ choices are usually justified by the momentum of the game in question. Specifically, what a team was successful in executing that day and what the opponent was able to dictate at will.

Applied Statistics in Fourth Down Play Calling

A good example of effective use of metrics is in offensive play calling on fourth down. Fourth down play calling analysis uses Expected Points to guide decisions. Expected points uses a numerical value to express the field position of every down and distance situation.

The further an offensive team goes into the opponent’s side of the field, the lower the net yardage value of punting. Punting from deeper inside your opponent’s side of the field (closer to their end zone) increases the possibility of a touchback thereby devaluing the field position value of the punting option.

In Expected Points, the opponent’s resulting field position if the fourth down is not converted is factored into decisions. In other words, a failed fourth down attempt could be more valuable than a punt in pinning an opponent deeper in their own territory.

Importantly, the Expected Points metric only measures the value of a converted fourth down assuming the exact number of yards needed for the first down are gained. As experienced fans know, successful fourth down conversions often exceed the exact number of yards needed for a conversion. This means that a converted fourth down that exceeds the exact number of yards needed is of even greater value than shown by the Expected Points metric.

The 37 yard line is No Man’s Land

Assume your team has a fourth down and three at the opponents 37 yard line, early in the second quarter with the game tied. How to determine the best play call?

Punt — A punt from the opponent’s 37 yard line averages a net distance of 23 yards, placing your opponent on their 14 yard line for an Expected Points Value of +0.2.

Field Goal — A Field Goal from the opponents 37 yard line would be successful 45% of the time for an Expected Point Value of +2.3 factoring in the opponents field position after the kickoff. A missed field goal from the 37 yard line would give the ball to the opponent at their own 44 yard line for an Expected Points value of -1.1. The mathematical value of the field goal attempt is expressed as:

(0.45 * 2.3) + ((1–0.45) * -1.1) = 0.4 Expected Points

Fourth Down Conversion — Fourth Down and three conversions from the opponent’s 37 yard line are successful 56% of the time. This play would produce an Expected Points Value of +3.0 assuming exactly three yards were gained. A failed conversion would give the opponent the ball at their own 37 yard line for am Expected Point Value of -1.1. The mathematical value of the fourth down conversion attempt is expressed as:

(0.56 * 3.0) + ((1–0.56) * -1.1) = 1.2 Expected Points

So from this we can deduce that, not accounting for crunch time, going for the fourth down conversion is the most sensible option, even if it fails.

The 37 yard line is known as “No Man’s Land” for the statistical favorability of choosing a fourth down conversion over a field goal or punt.

Statistics can express probabilities to guide situational decision making. Football is still a game of human beings and character traits. There is still no test that can quantify the innate emotional and leadership qualities that separate the champions from the contenders. The future will reveal just how close we get to an Artificial Intelligence play calling system that can make owners look like geniuses without ever having strapped on a helmet.

NFLSportsStatistical analysisStrategyFootball

Comments / 0

Published by

disparate parts coalescing toward a greater meaning in the pursuit of a fully realized life

471 followers

More from Stuart Grant

Comments / 0